JPH0953399A - Tunnel ventilation control device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve control performance, precisely keep VI value within an allowable range, and reduce the required power. SOLUTION: Jet fans 2-1 2-n, VI meter 3-1, 3-2, and an anemometer 4 are set within a tunnel 1. Fuzzy inference means 6 inputs measured value and target value for VI value and wind velocity value through input means 5 to perform a fuzzy operation, and outputs the number ▵NJF1 of jet fans to be increased and decreased to the present operating number of jet fans. Inverse normalization processing means 7 outputs a value ▵NJF in which the ▵NJF1 is inverse normalized. Output processing means 8 outputs a value in which ▵NJF is added to the present operating number NF to ventilator operating means 9 as control number NJFOUT. Fuzzy rule learning means 10A measures the ignition frequency of each element of a fuzzy rule matrix used by the fuzzy inference means 6, and corrects the content of the fuzzy rule matrix so as to perform an optimum control.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a ventilation control device for a tunnel, and more particularly to a tunnel for measuring a pollution concentration in a road tunnel and operating a jet fan to control the pollution concentration within an allowable range. The present invention relates to a ventilation control device.

[0002]

2. Description of the Related Art In a road tunnel, the exhaust gas of an automobile pollutes the inside of the tunnel. Therefore, mechanical ventilation is performed in a tunnel having a long length or a high traffic volume so that the pollution concentration is kept below an allowable value. There is.

There are various kinds of pollutants in the tunnel, but the soot concentration, which causes the visibility in the tunnel to be lowered, is mainly controlled. The soot concentration is generally measured as a VI value (VI: Visibility Index). The VI value is a value representing the light transmittance and is 10
The closer to 0%, the better the visibility and the lower the soot concentration. A device for measuring the VI value is a haze transmittance meter (hereinafter, appropriately referred to as VI meter).

The ventilation system of a road tunnel includes a vertical flow ventilation system and a lateral flow ventilation system, but in recent years, the vertical flow ventilation system has become the mainstream. In a vertical-flow ventilation type road tunnel, the road itself is regarded as a ventilation duct, and ventilation air is blown in the longitudinal direction of the road to dilute pollutants. In other words, it is a method of sucking in fresh air from one of the wells and discharging polluted air from the other of the wells. A jet fan installed on the ceiling of the roadway is used as a device for flowing the ventilation air.

As the conventional control method, feedback control based on the measured value of the VI meter is most often used. As a concrete method, for example, when a target area of a VI value is set, a management level of several steps is provided above and below the area, and when the VI value is below a management level lower than the target area, , If the number of operating ventilators is increased, and if the control level above the target area is exceeded, the number of operating ventilators is decreased to maintain the VI value within the target area. is there.

In the case of face-to-face traffic in a vertical flow ventilation type tunnel, since there is a vehicle traveling in the direction opposite to the ventilation direction, the wind speed in the roadway tends to decrease, and the VI value also decreases with a delay. Therefore, in some cases, a control method is adopted in which not only the VI value but also the wind speed in the roadway is incorporated as a feedback element.

As an example of the above-mentioned control system incorporating the VI value and the wind speed in the roadway as feedback elements,
There is a control system that uses a fuzzy rule matrix for the VI value and the wind speed value. This fuzzy control method is shown in FIG.
This will be described with reference to FIGS.

FIG. 10 shows the system configuration of an embodiment of this fuzzy control system. The tunnel 1 is a road tunnel of a longitudinal flow ventilation method of two-way passage, and there are vehicles traveling from the A well entrance to the B well entrance and vehicles traveling from the B well entrance to the A well entrance. .., 2-n are installed in the tunnel 1, and the inside of the tunnel 1 is ventilated by these jet fans. The direction of ventilation air flow is from the A well entrance to the B well entrance. As the sensor, a VI meter 3-1 that measures a VI value is used.
3-2 and an anemometer 4 for measuring the wind speed in the roadway are installed.

The input processing means 5 includes VI meters 3-1 and 3-2.
Input value, the measured value of the wind direction anemometer 4, and the preset VI target value and wind speed target value, and the VI deviation Δ
The VI and the wind speed deviation ΔVr are calculated. The fuzzy inference means 6 inputs the VI deviation ΔVI and the wind speed deviation ΔVr calculated by the input processing means 5, and calculates the correction amount _{ΔN JF1} of the number of jet fan operating units by fuzzy inference. The inverse normalization processing means 7 determines the correction amount ΔN.
_JF1 is to calculating the inverse normalized △ N _JF in inverse normalization coefficient S _njf. The output processing means 8 adjusts the number of jet fan operating units calculated by the inverse normalization processing means 7 _{ΔN JF}
Then, the control output N _JFOUT is determined by adding the operating number of jet fans at that time. A fuzzy control device A ₁ is configured by these means 5, 6, 7, and 8. Then, the ventilator operating means 9 makes the jet fans 2-1, ..., 2-n so that the number of operating jet fans becomes equal to the control output determined by the output processing means 8.
It outputs the operation / stop command to the.

Next, the operation of FIG. 1 will be described. The input processing means 5 measures the measured value of the VI meter and the measured value of the anemometer and V.
By inputting the I target value and the wind speed target value, the following equation (1),
The wind speed deviation ΔVr of the VI deviation ΔVI is calculated by the equation (2).

ΔVI = min (VI ₁ −VI _ref1 , VI ₂ −VI _ref2 ) (1) VI ₁ : measurement value of VI meter 3-1 [%] VI ₂ : measurement value of VI meter 3-2 [ %] VI _ref1 : Target value of VI ₁ [%] VI _ref2 : Target value of VI ₂ [%] ΔVr = Vr−Vr _ref (2) Vr: Measured wind speed in the roadway [m / s] Vr _ref : Wind speed target value in the road [m / s] The VI target values VI _ref1 , VI _ref2, and the wind speed target value Vr _ref are preset values.

Then, the fuzzy inference means 6 inputs the VI deviation ΔVI and the wind speed deviation ΔVr calculated by the input processing means 5, and calculates the correction amount _{ΔN JF1} of the number of operating jet fans by fuzzy inference. Details of the processing contents are as follows.

First, the VI deviation ΔVI and the wind speed deviation ΔVr are normalized by the following equation.

ΔVI ← ΔVI / S _VI (3) ΔVr ← ΔVr / S _Vr (4) S _VI : ΔVI scale factor [%] S _Vr : ΔVr scale factor [m / s ] Next, the normalized ΔVI and ΔVr are input, and the correction amount ΔN of the number of operating jet fans is fuzzy inferred.
_Ask for _JF1 . The membership functions for the input variables ΔVI and ΔVr are shown in FIG. 11, and the output variables _{ΔN JF1}
The membership function for is shown in FIG. The output variable _{ΔN JF1} is also normalized to the range of -1 to 1. In FIG. 11 and FIG. 12, the meaning of the label of the membership function is as follows.

NB: Negative Big, NM: Negative Medium, NS: Negative Small Z: Zero PS: Positive Small, PM: Positive Medium, PB: Positive Big As fuzzy rules, for example, the following rules are used.

[0016] If △ VI is NB and △ Vr is NB then △ N JF1 isPB. ... (5) If △ VI is Z and △ Vr is NB then △ N JF1 isPS. (6) If ΔVI is Z and ΔVr is Z then _{ΔN JF1} is Z. (7) If ΔVI is NS and ΔVr is PS then _{ΔN JF1} is Z. ... (8) If △ VI is PB and △ Vr is PB then △ N JF1 isNM. (9) Fuzzy rule 5 means "if the VI value is significantly lower than the target value and the wind speed is also significantly lower than the target value, the number of jet fans to be operated is increased." Further, the fuzzy rule 6 means "if the VI value is near the target value and the wind speed is considerably low, the operating number of jet fans is slightly increased." If the wind speed is low even if the VI value is close to the target value, the VI value is likely to decrease with a delay. Therefore, in Rule 6, the number of operating jet fans is increased in order to prevent the VI value from decreasing. Figure 1 shows a table of such fuzzy rules.
3.

Since the correction amount _{ΔN JF1} of the number of operating jet fans calculated by the fuzzy inference means 6 is a normalized value, the inverse normal processing means 7 uses the following equation to denormalize the value ΔN.
Performs _JF calculation. _{△ N JF ← S NJF · △} N JF1 ... (10) S NJF: △ inverse normalization coefficient [units] output processing section 8 of the N _JF1 the inference result of the fuzzy inference means 6 △ N
Input _JF and calculate the control output (the number of jet fans operating) by the following formula. N _JFOUT = N _JF + △ N _JF (11) N _JF : Actual jet fan operating number [unit] at the time of control output calculation N _JFOUT : Control output [unit] Finally, the ventilator operating means 9 outputs When the control output N _JFOUT is received from the processing means 8 and the number of jet fans operating is N
Jet fan 2-1 to 2-n so that it becomes a _JFOUT stand
The command to start / stop is output to.

[0018]

In the above conventional method, when the fuzzy rule matrix is optimally determined, the fuzzy rule is fired as shown in FIG. 14, the control performance is good, and the required power is kept low. You can

However, in the conventional method, since the control rule and the denormalization gain are not learned online, the following problems occur when the fuzzy rule matrix is not optimally determined. (1) Among the elements of the fuzzy rule matrix, there are rules that are too weak, so the same rule (excluding the zero rule) is continuously fired as shown in FIG. 16, the control performance deteriorates, and the power consumption also increases. Get higher (2) Among the elements of the fuzzy rule matrix, there are rules that are too weak, so that rules of both extremes are fired alternately as shown in FIG. 15, hunting occurs, control performance deteriorates, and power consumption increases. (3) The characteristics of the controlled object change, the fuzzy rule matrix is not optimal, and the above phenomena (1) and (2) occur.

The present invention has been made in view of the above circumstances, and it is possible to improve the control performance, maintain the VI value within the allowable range with high accuracy, and reduce the required electric power. Is intended to provide.

[0021]

As a means for solving the above-mentioned problems, the invention according to claim 1 has a plurality of jet fans in a tunnel, and a haze transmittance meter installed in this tunnel. And, in the tunnel ventilation control device that controls the number of operating jet fans based on the measured values of the wind direction and anemometer, enter the measured values of the fume permeability meter and the wind direction anemometer to the current operating number of the jet fans. On the other hand, the fuzzy inference means for deciding the operating number to be increased or decreased by using fuzzy inference, and the firing frequency of each element of the fuzzy rule matrix used by the fuzzy inference means are measured, and the contents of the fuzzy rule matrix are corrected. And a fuzzy rule learning means.

According to a second aspect of the present invention, in the first aspect of the invention, the fuzzy rule learning means measures a firing locus instead of the firing frequency.
The present invention is characterized by comprising an inverse normalization coefficient learning means for measuring the firing locus and correcting the inverse normalization coefficient for the output of the fuzzy inference means.

According to a third aspect of the present invention, in the first aspect of the invention, the fuzzy rule learning means measures a firing locus in addition to the firing frequency.
An inverse normalization coefficient learning means for correcting the inverse normalization coefficient for the output of the fuzzy inference means is provided, by measuring the firing frequency and the firing locus.

According to a fourth aspect of the present invention, in the first aspect of the present invention, the fuzzy inference means inputs the measured values of the haze transmittance meter and the anemometer to calculate a first inferred value. A first fuzzy inference means, and a second inference value is calculated by inputting the first inference value and the rate of change of the measurement value of the haze transmittance meter, and the second inference value is determined as the operating number to be increased or decreased. The first stage and the second stage fuzzy inference means are comprised of two-stage fuzzy inference means, and the firing frequency of each element of the fuzzy rule matrix used by the first-stage and second-stage fuzzy inference means is measured, and the contents of these fuzzy rule matrices are respectively corrected. The present invention is characterized in that first-stage and second-stage fuzzy rule learning means are provided.

According to a fifth aspect of the present invention, in the fourth aspect of the invention, the first-stage and second-stage fuzzy rule learning means measure a firing locus instead of the firing frequency. A second-stage denormalization coefficient learning means for measuring the firing locus of the second-stage fuzzy inference means and correcting the denormalization coefficient for the output thereof is provided.

According to a sixth aspect of the present invention, in the fourth aspect of the invention, the first-stage and second-stage fuzzy rule learning means measure the firing locus in addition to the firing frequency. A second-stage denormalization coefficient learning means for measuring a firing frequency and a firing locus of the second-stage fuzzy learning means and correcting a denormalization coefficient for an output thereof is provided.

[0027]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block configuration diagram according to an embodiment of the first invention. In this figure, a tunnel 1 is a road tunnel of a longitudinal flow ventilation system of a two-way traffic, and there are a vehicle traveling from the A well entrance to the B well entrance and a vehicle traveling from the B well entrance to the A well entrance. Tunnel 1
Inside, n jet fans 2-1, ..., 2-n are installed, and the tunnel 1 is provided by these jet fans.
The inside is ventilated. The direction of ventilation air is from A to B
It is the direction of the wellhead. Further, as the sensors, VI meters 3-1 and 3-2 for measuring the VI value and a wind direction anemometer 4 for measuring the wind speed in the roadway are installed.

The input processing means 5 comprises VI meters 3-1 and 3-2.
Input value, the measured value of the wind direction anemometer 4, and the preset VI target value and wind speed target value, and the VI deviation Δ
The VI and the wind speed deviation ΔVr are calculated. The fuzzy inference means 6 inputs the VI deviation ΔVI and the wind speed deviation ΔVr calculated by the input processing means 5, and calculates the inference result by fuzzy inference. The inverse normalization means 7 is
The fuzzy inference result obtained by the fuzzy inference means 6 is input and denormalized, and the correction amount of the jet fan operating number Δ
N _JF is calculated. The output processing means 8 determines the control output N _JFOUT by adding the correction amount _{ΔN JF} of the number of operating jet fans calculated by the inverse normalizing means 7 to the number of operating jet fans at that time. In the ventilator operating means 9, the output processing means 8 is the number of jet fans operating.
The operation / stop is output to the jet fans 2-1, 2-2, ..., 2-n so as to be equal to the control output determined in (1).

Then, the fuzzy rule learning means 10A
Is to measure the firing frequency of the fuzzy rule matrix used in the fuzzy inference means 6, input the measured firing frequency, and modify the fuzzy rule matrix by the AI rule every N hours.

Next, the operation of FIG. 1 will be described. However, the input processing means 5, the fuzzy inference means 6, the denormalization processing means 7, and the output processing means 8 perform the same operations as the contents described in the conventional art, and thus duplicated description will be omitted.

The fuzzy rule learning means 10A measures the firing frequency of the fuzzy rule matrix used in the fuzzy inference means 6 and corrects the fuzzy rule level by the AI rule as shown below. That is, first, the fuzzy rule matrix is classified into an N region and a P region as shown in FIG. 2, and the bias of the relative firing frequency of each region is examined, and then the firing frequency of each element is examined. Thereby, the fuzzy rule to be modified is determined, and the modification is performed according to the AI rule. However, the firing of the Z (zero) rule is M
If it exceeds AX% (for example, 60%), no correction is made and the Z (zero) rule is not modified. if “The firing frequency of the N region is high” and “The firing frequency of the P region is not high” then “Increase the level of the firing rule of the N region” (12) if “The firing frequency of the P region is high” and “The firing frequency of the N area is not high” then “Increase the level of the firing rule of the P area”… (13) if “The firing frequency of the Z area is low” then “The firing of the P area and the N area (14) The AI rule of (12) above is that “if the firing frequency of the entire N region in FIG. 2 is high and the firing frequency of the entire P region is not high, the rule firing is N. It is regarded as being biased toward the region, and the level of all firing rules in the N region is increased (for example, 1 level). " Here, for example, the following classification is used as the classification of the firing frequency of the antecedent part of the above equation. "High firing frequency in P region": The firing frequency in P region accounts for x% (for example, 40%) or more of the entire firing frequency. “High firing frequency in N region”: The firing frequency in N region accounts for x% (for example, 40%) or more of the total firing frequency. “High firing frequency in Z region”: The firing frequency in the Z region occupies x% (for example, 40%) or more with respect to the entire firing frequency. "The firing frequency in the P region is not high": The firing frequency in the P region is smaller than x% (for example, 40%) with respect to the entire firing frequency. "The firing frequency in the N region is not high": The firing frequency in the N region is smaller than x% (for example, 40%) with respect to the entire firing frequency. “The firing frequency in the Z region is not high”: The firing frequency in the Z region is smaller than x% (for example, 40%) with respect to the entire firing frequency.

The consequent part of the above equation is executed as follows. "Increase level": Increase the level of the specified rule on the fuzzy rule matrix by several levels (for example, one level). “Lower level”: The level of the specified rule on the fuzzy rule matrix is decreased by several levels (for example, one level).

Next, an embodiment of the second invention will be described with reference to FIG. The fuzzy rule learning means 10B measures the firing locus of the fuzzy rule matrix used in the fuzzy inference means 6 and corrects the fuzzy rule level according to the AI rule as described below. However, when the firing of the Z (zero) rule exceeds MAX% (for example, 60%), the correction is not performed, and the Z (zero) rule is not corrected. Here, the firing locus focuses on the firing history of the fuzzy rules and the firing distance which is the distance from the previous firing point to the present firing point on the fuzzy rule matrix as shown in FIG. The firing distance is calculated based on the firing distance (not shown) and the calculation means based on the values of ΔVI and ΔVr. if “The firing path is long” then “Lower the level of the rule that fired last time” (14) if “The same rule fires continuously” then “Increase the level of the rule that fires continuously” (15) The above-mentioned (14) AI rule means that "when the firing distance on the firing locus is long, the level of the rule fired last time is lowered." Further, the AI rule of (15) means “when the same rule is continuously fired a certain number of times (for example, 10 times) or more, the level of the continuously fired rule is increased.” Further, here, for example, the following classification is used as the classification of the firing trajectory of the antecedent part of the above equation. The consequent part of the above equation is the same as in the case of the embodiment of the first invention. "Ignition locus is long": The distance from the previous firing rule to the present firing rule is longer than I _l (for example, 0.8). "The total of the firing loci is long": The total of the firing distances in the firing locus is longer than _Ll (for example, 0.8). "Short firing path": The distance from the previous firing rule to the current firing rule is shorter than I _s (for example, 0.2). “Short firing path”: The total firing distance in the firing path is shorter than L _s (eg 0.2). “Same rule fires continuously”: Focusing on the firing history, the same rule fires a certain number of times (for example, 10 times) or more in succession.

The denormalization coefficient learning means 11B measures the firing locus of the fuzzy rule matrix used in the fuzzy inference means 6 and corrects the denormalization coefficient according to the AI rule as shown below. However, if the firing of the Z (zero) rule exceeds MAX% (for example, 60%), no correction is made. Here, the firing locus of the fuzzy rule matrix focuses on the firing history of the fuzzy rule and the firing distance, which is the distance from the firing point on the fuzzy rule matrix, as in the case of the fuzzy rule learning means 10B. if “Total firing trajectory is long” then “Decrease denormalization coefficient” (16) if “Total firing trajectory is short” then “Increase denormalization coefficient” (17) Above (16 The AI rule of () means that when the total firing distance in the firing locus is long, it is considered that the inverse normalization coefficient is too large and hunting is performed, and the inverse normalization coefficient is decreased. Further, the AI rule of (17) means “when the total firing distance in the firing locus is short, the denormalization coefficient is considered to be too small and the denormalization coefficient is increased”. Here, the classification of the firing locus in the antecedent part of the above equation is the same as in the case of the fuzzy rule learning means 10B. The consequent part of the above equation is executed as follows. “Increase inverse normalization coefficient”: Inverse normalization coefficient is ΔS _p
Only increase. “Decrease inverse normalization coefficient”: Inverse normalization coefficient ΔS _n
Only decrease.

Next, an embodiment of the third invention will be described with reference to FIG. The fuzzy rule learning means 10C measures the firing frequency and firing trajectory of the fuzzy rule matrix used in the fuzzy inference means 6, and corrects the fuzzy rule level according to the AI rule as shown below. However, the firing of the Z (zero) rule is MAX% (eg 60
%), No correction shall be made, and Z
(Zero) Rules shall not be modified. here,
The firing frequency of the fuzzy rule matrix is checked by the same method as in the first embodiment of the invention. As for the firing locus, attention is paid to the firing history and the firing distance as in the case of the second embodiment.

If “the firing frequency of the N region is high” and “the firing frequency of the P region is not high” then “raise the level of the firing rule of the N region” (18) if “the firing frequency of the P region Is high ”and“ the firing frequency of the N region is not high ”then“ raise the level of the firing rule of the P region ”… (19) if“ the firing frequency of the N region is high ”and“ the firing frequency of the P region ” Is high ”and“ The total firing trajectory is long ”then“ Lower the level of firing rules in the N area ”and“ Lower the level of firing rules in the P area ”(20) if“ Fire path ” Is long ”then“ Lower the level of the firing rule ”… (21) if“ The same rule is fired a certain number of times consecutively ”then“ Increase the level of the firing rule ”… (22) The AI rule of (18) above is “N in FIG. When the firing frequency of the entire region is high and the firing frequency of the entire P region is not high, it is considered that the rule firing is biased toward the N region, and the level of all firing rules of the N region is increased. ” Further, the AI rule of (21) means “when the firing distance on the firing locus is long, the level of the rule fired last time is lowered”. Here, the classification of the firing frequency of the antecedent of the above equation is the same as in the case of the embodiment of the first invention, and the classification of the firing locus is the same as the case of the embodiment of the second invention. . The consequent part of the above equation is also the same as that of the first embodiment of the invention.

The denormalization coefficient learning means 11C measures the firing frequency and firing trajectory of the fuzzy rule matrix used in the fuzzy inference means 6, and corrects the denormalization coefficient according to the AI rule as shown below. However, if the firing of the Z (zero) rule exceeds MAX% (for example, 60%), no correction is made. Here, the firing frequency of the fuzzy rule matrix is checked by the same method as in the first embodiment of the invention. As for the ignition locus, attention is paid to the ignition history and the ignition distance as in the case of the second embodiment.

If “the firing frequency of the N region is high” and “the firing frequency of the P region is high” and “the firing locus is long” then “the denormalization coefficient is reduced” (23) if “the firing of the N region is fired” "Frequent frequency" and "High frequency of firing in P region" and "Short firing trajectory" then "Increase denormalization coefficient" ... (24) The AI rule in (23) above says that the total firing trajectory is If it is long, it is considered that the denormalization coefficient is too large and hunting is performed, and only the denormalization coefficient is decreased. ” Further, the AI rule of (24) means “when the total of the firing loci is short, the denormalization coefficient is considered to be too small and the denormalization coefficient is increased”.
Also, here, the classification of the firing frequency of the antecedent part of the above equation is
This is similar to the case of the embodiment of the first invention, and the classification of the firing locus is similar to that of the embodiment of the second invention. The consequent part of the above equation is the same as in the case of the embodiment of the second invention.

Next, an embodiment of the fourth invention will be described with reference to FIG. The input processing means 5 calculates the VI deviation ΔVI and the wind speed deviation ΔVr by the method described in the prior art. The first-stage fuzzy inference means 12 inputs the VI deviation ΔVI and the wind speed deviation ΔVr calculated by the input processing means f5, and uses fuzzy inference to correct the jet fan operating quantity Δ.
Calculate N _JF1 . Details of the processing contents are as follows.

First, the VI deviation ΔVI and the wind speed deviation ΔVr
Is normalized by the following equation.

ΔVI ← ΔVI / S _VI (25) ΔVr ← ΔVr / S _Vr (26) S _VI : ΔVI scale factor [%] S _Vr : ΔVr scale factor [m / s Next, the normalized ΔVI and ΔVr are input, and the amount of correction of the jet fan operating number ΔN is calculated by fuzzy reasoning.
_Ask for _JF1 . The membership function for the input variables ΔVI and ΔVr is the same as in FIG. 11, and the membership function for the output variable _{ΔN JF1} is the same as in FIG. The output variable _{ΔN JF1} is also normalized to the range of -1 to 1. In FIG. 11 and FIG. 12, the meaning of the label of the membership function is as follows.

NB: Negative Big, NM: Negative Medium, NS: Negative Small Z: Zero PS: Positive Small, PM: Positive Medium, PB: Positive Big As fuzzy rules, for example, the following rules are used.

[0043] If △ VI is NB and △ Vr is NB then △ N JF1 isPB. ... (27) If △ VI is Z and △ Vr is NB then △ N JF1 isPS. (28) If ΔVI is Z and ΔVr is Z then _{ΔN JF1} is Z. (29) If ΔVI is NS and ΔVr is PS then _{ΔN JF1} is Z. ... (30) If △ VI is PB and △ Vr is PB then △ N JF1 isNM. (31) The fuzzy rule of (27) described above is that "if the VI value is considerably lower than the target value and the wind speed is also significantly lower than the target value, the number of jet fans to be operated is increased."
It means. In addition, the fuzzy rule of (28) is, "If the VI value is near the target value and the wind speed is considerably low,
Increase the number of jet fans in operation. It means "." If the wind speed is low even if the VI value is near the target value, V
Since there is a strong possibility that the I value will decrease with a delay, the operating number of jet fans is increased in order to prevent a decrease in VI in advance in the rule (28). Such a fuzzy rule is summarized in a table in FIG. 13 already shown.

The first stage fuzzy rule learning means 13A is
The firing frequency of the fuzzy rule matrix used in the first stage fuzzy inference means 12 is measured, and A
The level of the fuzzy rule is corrected by the I rule.
The firing frequency of the fuzzy rule matrix is examined by the same method as in the first embodiment of the invention, the fuzzy rule to be corrected is determined, and the fuzzy rule is corrected according to the AI rule. However, the firing of the Z (zero) rule is MAX% (eg 60
%), No correction shall be made, and Z
(Zero) Rules shall not be modified.

If “the firing frequency of the N region is high” and “the firing frequency of the P region is not high” then “raise the level of the firing rule of the N region” (32) if “the firing frequency of the P region Is high ”and“ the firing frequency of the N region is not high ”then“ raise the level of the firing rule of the P region ”… (33) if“ The firing frequency of the Z region is low ”then“ P region and N region (34) The AI rule of (32) described above is that if the firing frequency of the entire N region in Fig. 2 is high and the firing frequency of the entire P region is not high, the rule is It is considered that the firing is concentrated in the N area, and the level of all firing rules in the N area is raised. " Here, the classification of the firing frequency in the antecedent part of the above formula and the consequent part of the above formula are the same as in the case of the embodiment of the first invention.

The VI change rate calculating means 14 is a VI totalizer 3-1.
And the measured value of the VI meter 3-2 are input, and the rate of change of the VI value with time is calculated. The processing method is as follows.

The current measured values of the VI meter 3-1 and the VI meter 3-2 are compared, and the change rate is calculated for the smaller VI value by the following formula.

ΔVI / Δt = {VI (t) −VI (t−Δt)} / Δt (35) ΔVI / Δt: Change rate of VI [% / s] Δt: Time step [S] VI (t): current VI value [%] VI (t-Δt): VI value before Δt seconds [%] The second stage fuzzy inference means 15 is the first stage fuzzy inference means 12. Amount of correction of the calculated number of jet fans operating △
N _JF1 and the VI change rate _ΔVI / Δt calculated by the VI change rate calculating means 14 are input, and the final correction amount _{ΔN JF2} of the jet fan operating number is calculated by fuzzy inference. The processing procedure is as follows.

First, the VI change rate is normalized by the following equation.

ΔVI / Δt ← (ΔVI / Δt) / S _DVI (36) S _DVI : Scale factor for the VI change rate [% /
s] Note that the correction amount _{ΔN JF1} of the number of jet fan operating units has already been normalized by the first-stage fuzzy inference means 12, so normalization is not performed here.

Next, the correction amount _{ΔN JF1} and the VI change rate _ΔVI / Δt of the number of operating jet fans are used as input variables.
The final correction amount _{ΔN JF2} of the number of operating jet fans is calculated by fuzzy reasoning. The membership function for the VI change rate ΔVI / Δt is the same as the membership function for the VI deviation and the wind speed deviation shown in FIG. 11. Further, the membership function for the correction amount △ N _JF2 jet fan operating number is the output variable of the second stage fuzzy inference means 15 also members for output variable △ N _JF1 the first stage fuzzy inference means 12 shown in FIG. 12 It is the same as the ship function.

As the fuzzy rules, for example, the following rules are used.

[0053] _{If △ N JF1 isZand △ VI /} △ tisNSthen △ N JF2 isPS. _{... (37) If △ N JF1} isZand △ VI / △ tisZ then △ N JF2 isZ. _{... (38) If △ N JF1} isZand △ VI / △ tisPSthen △ N JF2 isNS. (39) In the case of the above rule (37), the output _{ΔN JF1} of the first-stage fuzzy inference means 12 is zero, but the VI value tends to decrease, so the number of jet fans operating is slightly increased. On the contrary, according to the rule (39), since the VI value tends to increase, the number of jet fan operating units is slightly decreased. FIG. 7 shows a table of the fuzzy rules used in the second-stage fuzzy inference means 15.

The second stage fuzzy rule learning means 16A is
The firing frequency of the fuzzy rule matrix used in the second stage fuzzy inference means 15 is measured, and A
The level of the fuzzy rule is corrected by the I rule.
Here, the firing frequency of the fuzzy rule matrix is the first
In the same manner as in the embodiment of the invention, the fuzzy rule to be modified is determined and the modification is performed according to the AI rule, but the Z (zero) rule is not modified.

If “the firing frequency of the N region is high” and “the firing frequency of the P region is not high” then “raise the level of the firing rule of the N region” (40) if “the firing frequency of the P region Is high ”and“ the firing frequency of the N area is not high ”then“ raise the level of the firing rules of the P area ”(41) if“ the firing frequency of the Z area is low ”then“ the P area and the N area (42) The AI rule of (40) above is the rule that "if the firing frequency of the entire N region in Fig. 2 is high and the firing frequency of the entire P region is not high, It is considered that the firing is concentrated in the N area, and the level of all firing rules in the N area is raised. " Here, the classification of the firing frequency of the antecedent part of the above formula and the consequent part of the above formula are the same as in the case of the embodiment of the first invention.

In the denormalization means 17, the correction amount _{ΔN JF2} of the number of operating jet fans calculated by the second-stage fuzzy inference means 15 is a normalized value, and thus the denormalized value _ΔN is obtained by the following equation. Performs _JF calculation.

_{ΔN JF} ← S _NJF · _{ΔN JF2} (43) S _NJF : _{Denormalization} coefficient of _{ΔN JF2} [unit] The output processing means 18 outputs the inference result _{ΔN JF2} of the second-stage fuzzy inference means 15 Input and calculate the control output (number of jet fans operating) by the following formula.

N _JFOUT = N _JF + _{ΔN JF} (44) N _JF : Actual jet fan operating number [unit] at the time of control output calculation N _JFOUT : Control output [unit] The ventilator operating means 9 is conventional. In the same method as the technique of (1), the command to operate / stop is output to the jet fans 2-1 to 2-n so that the number of operating jet fans becomes N _JFOUT units.

Next, an embodiment of the fifth invention will be described with reference to FIG. The input processing means 5 and the ventilator operating means 9 are the same as those described in the conventional art. First-stage fuzzy inference means 12, VI change rate calculation means 14, second-stage fuzzy inference means 15, denormalization processing means 17, output processing means 18
Are the same as the contents described in the embodiment of the fourth invention.

The first-stage fuzzy rule learning means 13B is
The firing trajectory of the fuzzy rule matrix used in the first-stage fuzzy inference means 12 is measured, and the following A
The level of the fuzzy rule is corrected by the I rule.
However, the firing of the Z (zero) rule is MAX% (for example, 6
If it exceeds 0%, no correction shall be made, and
The Z (zero) rule will not be modified. Here, the firing locus is the same as in the case of the embodiment of the second invention,
Focus on the firing history and firing distance.

If “Long firing trajectory” then “Lower level of rule fired last time” (45) if “Same rule fires continuously” then “Increase level of rule fired continuously”… (46) The AI rule of (45) described above means "if the firing distance on the firing locus is long, lower the level of the rule fired last time." Further, the AI rule (46) means “when the same rule is continuously fired a certain number of times (for example, 10 times) or more, the level of the continuously fired rule is increased”. Here, the classification of the firing locus of the antecedent part of the above formula and the consequent part of the above formula are the same as in the case of the embodiment of the second invention.

The second stage fuzzy rule learning means 16B is
The firing trajectory of the fuzzy rule matrix used in the second stage fuzzy inference means is measured, and the level of the fuzzy rule is corrected by the AI rule. However, if the firing of the Z (zero) rule exceeds MAX% (for example, 60%),
No modification shall be made, and the Z (zero) rule shall not be modified. Here, the ignition locus is focused on the ignition history and the ignition distance, as in the case of the embodiment of the second invention.

If “The firing locus is long” then “Lower the level of the rule that fired last time” (47) if “The same rule fires continuously” then “Increase the level of the rule fired continuously”… (48) The AI rule of (45) described above means "if the firing distance on the firing locus is long, lower the level of the rule fired last time." Further, the AI rule (46) means “when the same rule is continuously fired a certain number of times (for example, 10 times) or more, the level of the continuously fired rule is increased”. Here, the classification of the firing locus of the antecedent part of the above formula and the consequent part of the above formula are the same as in the case of the embodiment of the second invention.

The second-stage denormalization coefficient learning means 19B measures the firing locus of the fuzzy rule matrix used in the second-stage fuzzy inference means 15, and the AI as shown below is obtained.
Correct the denormalization coefficient according to the rule. However, Z
If the firing of the (zero) rule exceeds MAX% (eg 60%), no correction is made. Here, the firing locus of the fuzzy rule matrix is focused on the firing history and the firing distance as in the case of the embodiment of the second invention.

If “Total firing trajectory is long” then “Decrease denormalization coefficient” (49) if “Total firing trajectory is short” then “Increase denormalization coefficient” (50) The (49) AI rule means “when the total firing distance in the firing locus is long, it is considered that the inverse normalization coefficient is too large and hunting is performed, and the inverse normalization coefficient is decreased”. Further, the AI rule (50) means “when the total firing distance in the firing locus is short, the denormalization coefficient is considered to be too small and the denormalization coefficient is increased”. Here, the classification of the firing locus of the antecedent part of the above formula and the consequent part of the above formula are the same as in the case of the embodiment of the second invention.

Next, an embodiment of the sixth invention will be described with reference to FIG. The input processing means 5 and the ventilator operating means 9 are the same as those described in the conventional art. First-stage fuzzy inference means 12, VI change rate calculation means 14, second-stage fuzzy inference means 15, denormalization processing means 17, output processing means 18
Are the same as the contents described in the embodiment of the fourth invention.

The first-stage fuzzy rule learning means 13C is
The firing frequency and firing trajectory of the fuzzy rule matrix used in the first-stage fuzzy inference means 12 are measured, and the fuzzy rule level is corrected by the AI rule as shown below. However, ignition of Z (zero) rule is MAX%
If it exceeds (for example, 60%), the correction is not performed, and the Z (zero) rule is not corrected. Here, the firing locus of the fuzzy rule matrix is examined by the same method as in the first embodiment of the invention. As for the ignition locus, attention is paid to the ignition history and the ignition distance as in the case of the second embodiment.

If “the firing frequency of the N region is high” and “the firing frequency of the P region is not high” then “raise the level of the firing rule of the N region” (51) if “the firing frequency of the P region” Is high ”and“ the firing frequency of the N region is not high ”then“ raise the level of the firing rule of the P region ”… (52) if“ the firing frequency of the N region is high ”and“ the firing frequency of the P region ” Is high ”and“ the total firing locus is long ”then“ Lower the level of firing rules in N area ”and“ Lower the level of firing rules in P area ”(53) if“ Ignition locus ” Is long ”then“ Lower the level of the firing rule ”… (54) if“ The same rule is fired a certain number of times consecutively ”then“ Increase the level of the firing rule ”… (55) The AI rule of (51) above is “N in FIG. When the firing frequency of the entire region is high and the firing frequency of the entire P region is not high, it is considered that the rule firing is biased toward the N region, and the level of the firing rule of the N region is increased. ” Further, the AI rule of (54) means "when the firing distance on the firing locus is long, the level of the rule fired last time is lowered". Here, the classification of the firing frequency of the antecedent part of the above equation is the same as in the case of the embodiment of the first invention, and the classification of the firing locus is the same as the case of the embodiment of the second invention. The consequent part of the above equation is the same as that of the first embodiment of the invention.

The second stage fuzzy rule learning means 16C is
The firing frequency and firing trajectory of the fuzzy rule matrix used in the second-stage fuzzy inference means 15 are measured, and the fuzzy rule level is corrected by the AI rule as shown below. However, ignition of Z (zero) rule is MAX%
If it exceeds (for example, 60%), the correction is not performed, and the Z (zero) rule is not corrected. Here, the firing frequency of the fuzzy rule matrix is checked by the same method as in the first embodiment of the invention. As for the ignition locus, attention is paid to the ignition history and the ignition distance as in the case of the second embodiment.

If “the firing frequency of the N region is high” and “the firing frequency of the P region is not high” then “increase the level of the firing rule of the N region” (56) if “the firing frequency of the P region” Is high ”and“ the firing frequency of the N region is not high ”then“ raise the level of the firing rule of the P region ”… (57) if“ the firing frequency of the N region is high ”and“ the firing frequency of the P region ” Is high ”and“ the total firing locus is long ”then“ Lower the level of firing rules in N area ”and“ Lower the level of firing rules in P area ”(58) if“ Ignition locus ” Is long ”then“ Lower the level of the firing rule ”… (59) if“ The same rule is fired a certain number of times consecutively ”then“ Increase the level of the firing rule ”… (60) The AI rule of (56) above is “N in FIG. If the firing frequency of the entire region is high and the firing frequency of the entire P region is not high, it is considered that the rule firing is biased toward the N region, and all levels of the firing rules of the N region are raised. ”
It means. Further, the AI rule of (59) means “when the firing distance on the firing locus is long, the level of the rule fired last time is lowered”. Here, the classification of the firing frequency of the antecedent part of the above equation is the same as in the case of the embodiment of the first invention, and the classification of the firing locus is the same as the case of the embodiment of the second invention. The consequent part of the above equation is the same as that of the first embodiment of the invention.

The second-stage denormalization coefficient learning means 19C measures the firing frequency and firing locus of the fuzzy rule matrix used in the second-stage fuzzy inference means 15, and performs denormalization by the AI rule as shown below. Correct the coefficient. However, the firing of the Z (zero) rule is MAX% (eg 60
%) Is exceeded, no correction is made. Here, the firing frequency of the fuzzy rule matrix is checked by the same method as in the case of the embodiment of the first invention. As for the ignition locus, attention is paid to the ignition history and the ignition distance as in the case of the second embodiment.

If “Ignition frequency in N area is high” and “Ignition frequency in P area is high” and “Ignition locus is long” then “Denormalization coefficient is reduced” (61) if “Ignition frequency in N area” "Frequent frequency" and "High frequency of firing in P region" and "Short firing trajectory" then "Increase denormalization coefficient" (62) The AI rule in (61) above says that the total firing trajectory is If it is long, it is considered that the inverse normalization coefficient is too large and hunting is performed, and the inverse normalization coefficient is decreased. ” In addition, the AI rule of (62) states that "when the total of the firing loci is short, the denormalization coefficient is considered to be too small,
Increase the denormalization factor. It means "." Here, the classification of the firing frequency in the antecedent part of the above equation is the same as in the case of the embodiment of the first invention, and the classification of the ignition locus and the consequent part of the above equation are in the embodiment of the second invention. It is similar to the case.

According to the embodiments of the present invention described above, the following various effects are achieved. That is, according to the embodiment of the first invention, in the tunnel ventilation control device using the fuzzy control, focusing on the firing frequency of the fuzzy rules and learning the fuzzy rule matrix, it is possible to avoid unnecessary fuzzy rules. It is possible to reduce movements, perform efficient control, efficiently maintain the VI value within an allowable range, and reduce power consumption.

According to the second embodiment of the present invention, in a tunnel ventilation control device using fuzzy control, it is possible to learn the fuzzy rule matrix and the denormalization coefficient by paying attention to the firing trajectory of the fuzzy rule. Therefore, unnecessary movements due to improper fuzzy rules and denormalization coefficients are reduced, and efficient control becomes possible.
The I value can be efficiently maintained within the allowable range, and the power consumption can be reduced.

According to the third embodiment of the present invention, in the tunnel ventilation control device using the fuzzy control, focusing on the firing frequency and the firing locus of the fuzzy rule, the fuzzy rule matrix and the denormalization coefficient are learned. Higher precision learning is possible, unnecessary movements due to inappropriate fuzzy rules and denormalization coefficients are reduced, efficient control is possible, VI value can be efficiently maintained within the allowable range, and consumption is also possible. Electric power can be reduced.

According to the fourth embodiment of the invention, in the tunnel ventilation control device using the two-stage inference type fuzzy control, paying attention to the firing frequency of the first-stage fuzzy rule, the first-stage fuzzy rule matrix is Focusing on the firing frequency of the second-stage fuzzy rules, learning the second-stage fuzzy rule matrix reduces unnecessary movements due to inappropriate fuzzy rules, enables efficient control, and allows VI values. High efficiency and high accuracy can be maintained within the range, and power consumption can be reduced.

According to the fifth embodiment of the present invention, in the tunnel ventilation control device using the two-stage inference type fuzzy control, focusing on the firing locus of the first-stage fuzzy rule, the first-stage fuzzy rule matrix is Focusing on the firing locus of the second-stage fuzzy rule, learning the second-stage fuzzy rule and the denormalization coefficient reduces unnecessary movements due to inappropriate fuzzy rules and the denormalization coefficient, resulting in efficient control. Therefore, the VI value can be maintained within the allowable range with high efficiency and high accuracy, and the power consumption can be reduced.

According to the sixth embodiment of the present invention, in the tunnel ventilation control apparatus using the two-stage inference type fuzzy control, the first stage fuzzy rule matrix is focused on by paying attention to the firing frequency and firing locus of the first stage fuzzy rule. Also, by paying attention to the firing frequency and firing trajectory of the second-stage fuzzy rule and learning the second-stage fuzzy rule and the denormalization coefficient, more accurate learning becomes possible, and inappropriate fuzzy rules and inverse It is possible to reduce wasteful movements due to the normalization coefficient, enable efficient control, maintain the VI value within a permissible range with high efficiency and high accuracy, and reduce power consumption.

Although the denormalization coefficient learning means is not provided in the configurations of the embodiments of the first and fourth inventions (FIGS. 1 and 6), the denormalization coefficient learning means is used only for the firing frequency. This is because it is not possible to obtain enough information to correct the. For example, if the N region of the fuzzy rule matrix ignites 10 times in a row and then the P region ignites 10 times in a row (when correction is required), the N region and the P region ignite alternately 20 times. It is not possible to distinguish from the case (when correction is unnecessary) only by the firing frequency.

[0080]

As described above, according to the present invention, it is possible to improve the control performance, maintain the VI value within the allowable range with high accuracy, and reduce the required power.

[Brief description of drawings]

FIG. 1 is a block configuration diagram according to an embodiment of a first invention.

FIG. 2 is an explanatory diagram showing the contents of a fuzzy rule matrix used by the fuzzy inference means in FIG.

FIG. 3 is a block configuration diagram according to an embodiment of the second invention.

FIG. 4 is an explanatory diagram of a firing trajectory of a fuzzy rule matrix used by the fuzzy inference means in FIG.

FIG. 5 is a block configuration diagram according to an embodiment of the third invention.

FIG. 6 is a block configuration diagram according to an embodiment of a fourth invention.

FIG. 7 is an explanatory diagram showing the contents of a fuzzy rule matrix used by the second-stage fuzzy inference means in FIG.

FIG. 8 is a block configuration diagram according to an embodiment of the fifth invention.

FIG. 9 is a block configuration diagram according to an embodiment of a sixth invention.

FIG. 10 is a block configuration diagram according to a conventional example.

FIG. 11 is a characteristic diagram showing a membership function for an input variable of fuzzy inference in a conventional example and the present invention.

FIG. 12 is a characteristic diagram showing a membership function for an output variable of fuzzy inference in a conventional example and the present invention.

FIG. 13 is a characteristic diagram showing a membership function for an output variable of fuzzy inference in a conventional example and the present invention.

FIG. 14 is an explanatory diagram showing a firing situation of a fuzzy rule matrix of fuzzy inference in a conventional example.

FIG. 15 is an explanatory diagram showing a firing state of a fuzzy rule matrix for fuzzy inference in a conventional example.

FIG. 16 is an explanatory diagram showing a firing situation of a fuzzy rule matrix of fuzzy inference in a conventional example.

[Explanation of symbols]

1 tunnel 2-1 to 2-n jet fan 3-1 and 3-2 VI meter (fume permeability meter) 4 wind direction anemometer 5 input processing means 6 fuzzy inference means 7,17 denormalization processing means 8,18 output Processing means 9 Ventilator operating means 10A, 10B, 10C Fuzzy rule learning means 11B, 11C Denormalization coefficient learning means 12 First stage fuzzy inference means 13A, 13B, 13C First stage fuzzy rule learning means 14 VI Change rate computing means 15 Second-stage fuzzy inference means 16A, 16B, 16C Second-stage fuzzy rule learning means 19B, 19C Second-stage denormalization coefficient learning means

Claims (6)

[Claims] 1. A tunnel ventilation control which has a plurality of jet fans in a tunnel and controls the number of jet fans operating based on the measured values of a haze transmittance meter and anemometer installed in the tunnel. In the device, by inputting the measurement values of the haze transmittance meter and the wind direction anemometer, the number of operating machines that should be increased or decreased with respect to the current operating number of the jet fan, fuzzy inference means for determining by using fuzzy inference, and A tunnel ventilation control device comprising: a fuzzy rule learning means for measuring the firing frequency of each element of the fuzzy rule matrix used by the fuzzy inference means, and correcting the contents of the fuzzy rule matrix. 2. The tunnel ventilation control device according to claim 1, wherein the fuzzy rule learning means measures an ignition locus instead of the firing frequency, and further measures the ignition locus to obtain the fuzzy rule. A tunnel ventilation control device comprising denormalization coefficient learning means for correcting the denormalization coefficient for the output of the inference means. 3. The tunnel ventilation control device according to claim 1, wherein the fuzzy rule learning means measures a firing locus in addition to the firing frequency, and further measures the firing frequency and the firing locus. A tunnel ventilation control device comprising: a denormalization coefficient learning unit that corrects a denormalization coefficient for the output of the fuzzy inference unit. 4. The tunnel ventilation control device according to claim 1, wherein the fuzzy inference means inputs the measured values of the haze transmittance meter and the wind direction anemometer to calculate a first inferred value. Means, and a second-stage fuzzy for inputting the first inference value and the rate of change of the measurement value of the haze transmittance meter to calculate a second inference value, and determining this as the operating number to be increased or decreased. A first stage and a second stage for measuring the firing frequency of each element of the fuzzy rule matrix used by the first-stage and second-stage fuzzy inference units, and correcting the contents of these fuzzy rule matrices. A tunnel ventilation control device comprising a second-stage fuzzy rule learning means. 5. The tunnel ventilation control device according to claim 4, wherein the first-stage and second-stage fuzzy rule learning means measures an ignition locus instead of the ignition frequency, and further, the second aspect. A tunnel ventilation control device comprising: a second stage denormalization coefficient learning means for measuring an ignition trajectory of the stage fuzzy inference means and correcting an inverse normalization coefficient for the output thereof. 6. The tunnel ventilation control device according to claim 4, wherein the first-stage and second-stage fuzzy rule learning means measure an ignition locus in addition to the ignition frequency. A tunnel ventilation control device comprising: a second-stage denormalization coefficient learning unit that measures a firing frequency and a firing locus of the stage fuzzy learning unit and corrects a denormalization factor for an output thereof.